Qualcomm Announces AWS Collaboration on Custom AI Inference Chips

Qualcomm has announced a collaboration with AWS covering multiple generations of custom chips for AWS AI infrastructure. The stated focus is AI inference, rather than a disclosed training deployment.

The announcement provides no details on the chip models involved, expected supply volumes, infrastructure capacity, implementation schedule, commercial terms or the AWS geographies that may be covered. As a result, the agreement should be viewed as a disclosed technology and infrastructure collaboration rather than evidence of a defined hardware order or a committed data-center build-out.

For hyperscale operators, the arrangement points to a potential broadening of the hardware base supporting AI data centers. While GPUs remain central to many AI training environments, custom and energy-efficient CPU or system-level solutions may assume greater importance for inference workloads. This is an inference from the announced focus, not a disclosed deployment outcome.

The collaboration is Qualcomm’s second disclosed major hyperscaler agreement following its previously reported relationship with Meta. According to S1, Meta is expected to use Qualcomm’s Dragonfly C1000 processor after it enters mass production in 2028.

Qualcomm has previously set a target of US$15 billion in data-center business revenue for fiscal 2029. That figure is a corporate target, not reported revenue. Separately, S1 cited a Bank of America forecast that the data-center CPU market will grow from US$27 billion in 2025 to US$60 billion in 2030; these are forecasts rather than observed market values.